Improving the Annotation for Spatial Proteomics: A Computational Approach to Enhance Molecular Characterization of...

Area of Science:

• Biomedical Imaging
• Proteomics
• Computational Pathology

Background:

• Matrix-assisted laser-desorption ionization mass spectrometry imaging (MALDI-MSI) shows potential for molecular characterization of thyroid neoplasms.
• Challenges in MALDI-MSI include minimizing signal interference and enhancing diagnostic discrimination.

Purpose of the Study:

• To develop a reproducible, automated workflow integrating digital pathology with MALDI-MSI for thyroid tissue analysis.
• To enhance spatial proteomics and diagnostic precision in thyroid neoplasms.

Main Methods:

• Developed a pixel classifier for automated selection of cell-rich regions of interest (ROIs) from H&E-stained thyroid tissue microarrays.
• Compared proteomics signals from pixel classifier (PC) selected ROIs against full core (FC) and pathologist (PAT) annotations.
• Utilized principal component analysis and receiver operating characteristic (ROC) analysis for data interpretation.

Main Results:

• Pixel classifier ROIs reduced interfering signals by 15% and increased tryptic peptide signal-to-noise ratio by 37%.
• Detected 9-24% more m/z signals, improving spectral clustering and distinguishing histopathological regions.
• Achieved a 50% increase in discriminatory m/z features for thyroid nodule diagnosis compared to FC and PAT data.

Conclusions:

• The automated pixel classifier workflow enhances MALDI-MSI reproducibility and reduces operator workload.
• This approach optimizes MALDI-MSI for molecular characterization of thyroid neoplasms.
• The method holds potential for improved biomarker discovery and diagnostic precision in clinical pathology.